Archipelago » qemu

/*

 * Block driver for the QCOW version 2 format

 *

 * Copyright (c) 2004-2006 Fabrice Bellard

 *

 * Permission is hereby granted, free of charge, to any person obtaining a copy

 * of this software and associated documentation files (the "Software"), to deal

 * in the Software without restriction, including without limitation the rights

 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell

 * copies of the Software, and to permit persons to whom the Software is

 * furnished to do so, subject to the following conditions:

 *

 * The above copyright notice and this permission notice shall be included in

 * all copies or substantial portions of the Software.

 *

 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR

 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,

 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL

 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER

 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,

 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN

 * THE SOFTWARE.

 */

#include "qemu-common.h"

#include "block/block_int.h"

#include "qemu/module.h"

#include <zlib.h>

#include "qemu/aes.h"

#include "block/qcow2.h"

#include "qemu/error-report.h"

#include "qapi/qmp/qerror.h"

#include "qapi/qmp/qbool.h"

#include "trace.h"

/*

  Differences with QCOW:

  - Support for multiple incremental snapshots.

  - Memory management by reference counts.

  - Clusters which have a reference count of one have the bit

    QCOW_OFLAG_COPIED to optimize write performance.

  - Size of compressed clusters is stored in sectors to reduce bit usage

    in the cluster offsets.

  - Support for storing additional data (such as the VM state) in the

    snapshots.

  - If a backing store is used, the cluster size is not constrained

    (could be backported to QCOW).

  - L2 tables have always a size of one cluster.

*/

typedef struct {

    uint32_t magic;

    uint32_t len;

} QCowExtension;

#define  QCOW2_EXT_MAGIC_END 0

#define  QCOW2_EXT_MAGIC_BACKING_FORMAT 0xE2792ACA

#define  QCOW2_EXT_MAGIC_FEATURE_TABLE 0x6803f857

static int qcow2_probe(const uint8_t *buf, int buf_size, const char *filename)

{

    const QCowHeader *cow_header = (const void *)buf;

    if (buf_size >= sizeof(QCowHeader) &&

        be32_to_cpu(cow_header->magic) == QCOW_MAGIC &&

        be32_to_cpu(cow_header->version) >= 2)

        return 100;

    else

        return 0;

}

/* 

 * read qcow2 extension and fill bs

 * start reading from start_offset

 * finish reading upon magic of value 0 or when end_offset reached

 * unknown magic is skipped (future extension this version knows nothing about)

 * return 0 upon success, non-0 otherwise

 */

static int qcow2_read_extensions(BlockDriverState *bs, uint64_t start_offset,

                                 uint64_t end_offset, void **p_feature_table)

{

    BDRVQcowState *s = bs->opaque;

    QCowExtension ext;

    uint64_t offset;

    int ret;

#ifdef DEBUG_EXT

    printf("qcow2_read_extensions: start=%ld end=%ld\n", start_offset, end_offset);

#endif

    offset = start_offset;

    while (offset < end_offset) {

#ifdef DEBUG_EXT

        /* Sanity check */

        if (offset > s->cluster_size)

            printf("qcow2_read_extension: suspicious offset %lu\n", offset);

        printf("attempting to read extended header in offset %lu\n", offset);

#endif

        if (bdrv_pread(bs->file, offset, &ext, sizeof(ext)) != sizeof(ext)) {

            fprintf(stderr, "qcow2_read_extension: ERROR: "

                    "pread fail from offset %" PRIu64 "\n",

                    offset);

            return 1;

        }

        be32_to_cpus(&ext.magic);

        be32_to_cpus(&ext.len);

        offset += sizeof(ext);

#ifdef DEBUG_EXT

        printf("ext.magic = 0x%x\n", ext.magic);

#endif

        if (ext.len > end_offset - offset) {

            error_report("Header extension too large");

            return -EINVAL;

        }

        switch (ext.magic) {

        case QCOW2_EXT_MAGIC_END:

            return 0;

        case QCOW2_EXT_MAGIC_BACKING_FORMAT:

            if (ext.len >= sizeof(bs->backing_format)) {

                fprintf(stderr, "ERROR: ext_backing_format: len=%u too large"

                        " (>=%zu)\n",

                        ext.len, sizeof(bs->backing_format));

                return 2;

            }

            if (bdrv_pread(bs->file, offset , bs->backing_format,

                           ext.len) != ext.len)

                return 3;

            bs->backing_format[ext.len] = '\0';

#ifdef DEBUG_EXT

            printf("Qcow2: Got format extension %s\n", bs->backing_format);

#endif

            break;

        case QCOW2_EXT_MAGIC_FEATURE_TABLE:

            if (p_feature_table != NULL) {

                void* feature_table = g_malloc0(ext.len + 2 * sizeof(Qcow2Feature));

                ret = bdrv_pread(bs->file, offset , feature_table, ext.len);

                if (ret < 0) {

                    return ret;

                }

                *p_feature_table = feature_table;

            }

            break;

        default:

            /* unknown magic - save it in case we need to rewrite the header */

            {

                Qcow2UnknownHeaderExtension *uext;

                uext = g_malloc0(sizeof(*uext)  + ext.len);

                uext->magic = ext.magic;

                uext->len = ext.len;

                QLIST_INSERT_HEAD(&s->unknown_header_ext, uext, next);

                ret = bdrv_pread(bs->file, offset , uext->data, uext->len);

                if (ret < 0) {

                    return ret;

                }

            }

            break;

        }

        offset += ((ext.len + 7) & ~7);

    }

    return 0;

}

static void cleanup_unknown_header_ext(BlockDriverState *bs)

{

    BDRVQcowState *s = bs->opaque;

    Qcow2UnknownHeaderExtension *uext, *next;

    QLIST_FOREACH_SAFE(uext, &s->unknown_header_ext, next, next) {

        QLIST_REMOVE(uext, next);

        g_free(uext);

    }

}

static void GCC_FMT_ATTR(2, 3) report_unsupported(BlockDriverState *bs,

    const char *fmt, ...)

{

    char msg[64];

    va_list ap;

    va_start(ap, fmt);

    vsnprintf(msg, sizeof(msg), fmt, ap);

    va_end(ap);

    qerror_report(QERR_UNKNOWN_BLOCK_FORMAT_FEATURE,

        bs->device_name, "qcow2", msg);

}

static void report_unsupported_feature(BlockDriverState *bs,

    Qcow2Feature *table, uint64_t mask)

{

    while (table && table->name[0] != '\0') {

        if (table->type == QCOW2_FEAT_TYPE_INCOMPATIBLE) {

            if (mask & (1 << table->bit)) {

                report_unsupported(bs, "%.46s",table->name);

                mask &= ~(1 << table->bit);

            }

        }

        table++;

    }

    if (mask) {

        report_unsupported(bs, "Unknown incompatible feature: %" PRIx64, mask);

    }

}

/*

 * Sets the dirty bit and flushes afterwards if necessary.

 *

 * The incompatible_features bit is only set if the image file header was

 * updated successfully.  Therefore it is not required to check the return

 * value of this function.

 */

int qcow2_mark_dirty(BlockDriverState *bs)

{

    BDRVQcowState *s = bs->opaque;

    uint64_t val;

    int ret;

    assert(s->qcow_version >= 3);

    if (s->incompatible_features & QCOW2_INCOMPAT_DIRTY) {

        return 0; /* already dirty */

    }

    val = cpu_to_be64(s->incompatible_features | QCOW2_INCOMPAT_DIRTY);

    ret = bdrv_pwrite(bs->file, offsetof(QCowHeader, incompatible_features),

                      &val, sizeof(val));

    if (ret < 0) {

        return ret;

    }

    ret = bdrv_flush(bs->file);

    if (ret < 0) {

        return ret;

    }

    /* Only treat image as dirty if the header was updated successfully */

    s->incompatible_features |= QCOW2_INCOMPAT_DIRTY;

    return 0;

}

/*

 * Clears the dirty bit and flushes before if necessary.  Only call this

 * function when there are no pending requests, it does not guard against

 * concurrent requests dirtying the image.

 */

static int qcow2_mark_clean(BlockDriverState *bs)

{

    BDRVQcowState *s = bs->opaque;

    if (s->incompatible_features & QCOW2_INCOMPAT_DIRTY) {

        int ret = bdrv_flush(bs);

        if (ret < 0) {

            return ret;

        }

        s->incompatible_features &= ~QCOW2_INCOMPAT_DIRTY;

        return qcow2_update_header(bs);

    }

    return 0;

}

static int qcow2_check(BlockDriverState *bs, BdrvCheckResult *result,

                       BdrvCheckMode fix)

{

    int ret = qcow2_check_refcounts(bs, result, fix);

    if (ret < 0) {

        return ret;

    }

    if (fix && result->check_errors == 0 && result->corruptions == 0) {

        return qcow2_mark_clean(bs);

    }

    return ret;

}

static QemuOptsList qcow2_runtime_opts = {

    .name = "qcow2",

    .head = QTAILQ_HEAD_INITIALIZER(qcow2_runtime_opts.head),

    .desc = {

        {

            .name = "lazy_refcounts",

            .type = QEMU_OPT_BOOL,

            .help = "Postpone refcount updates",

        },

        { /* end of list */ }

    },

};

static int qcow2_open(BlockDriverState *bs, QDict *options, int flags)

{

    BDRVQcowState *s = bs->opaque;

    int len, i, ret = 0;

    QCowHeader header;

    QemuOpts *opts;

    Error *local_err = NULL;

    uint64_t ext_end;

    ret = bdrv_pread(bs->file, 0, &header, sizeof(header));

    if (ret < 0) {

        goto fail;

    }

    be32_to_cpus(&header.magic);

    be32_to_cpus(&header.version);

    be64_to_cpus(&header.backing_file_offset);

    be32_to_cpus(&header.backing_file_size);

    be64_to_cpus(&header.size);

    be32_to_cpus(&header.cluster_bits);

    be32_to_cpus(&header.crypt_method);

    be64_to_cpus(&header.l1_table_offset);

    be32_to_cpus(&header.l1_size);

    be64_to_cpus(&header.refcount_table_offset);

    be32_to_cpus(&header.refcount_table_clusters);

    be64_to_cpus(&header.snapshots_offset);

    be32_to_cpus(&header.nb_snapshots);

    if (header.magic != QCOW_MAGIC) {

        ret = -EMEDIUMTYPE;

        goto fail;

    }

    if (header.version < 2 || header.version > 3) {

        report_unsupported(bs, "QCOW version %d", header.version);

        ret = -ENOTSUP;

        goto fail;

    }

    s->qcow_version = header.version;

    /* Initialise version 3 header fields */

    if (header.version == 2) {

        header.incompatible_features    = 0;

        header.compatible_features      = 0;

        header.autoclear_features       = 0;

        header.refcount_order           = 4;

        header.header_length            = 72;

    } else {

        be64_to_cpus(&header.incompatible_features);

        be64_to_cpus(&header.compatible_features);

        be64_to_cpus(&header.autoclear_features);

        be32_to_cpus(&header.refcount_order);

        be32_to_cpus(&header.header_length);

    }

    if (header.header_length > sizeof(header)) {

        s->unknown_header_fields_size = header.header_length - sizeof(header);

        s->unknown_header_fields = g_malloc(s->unknown_header_fields_size);

        ret = bdrv_pread(bs->file, sizeof(header), s->unknown_header_fields,

                         s->unknown_header_fields_size);

        if (ret < 0) {

            goto fail;

        }

    }

    if (header.backing_file_offset) {

        ext_end = header.backing_file_offset;

    } else {

        ext_end = 1 << header.cluster_bits;

    }

    /* Handle feature bits */

    s->incompatible_features    = header.incompatible_features;

    s->compatible_features      = header.compatible_features;

    s->autoclear_features       = header.autoclear_features;

    if (s->incompatible_features & ~QCOW2_INCOMPAT_MASK) {

        void *feature_table = NULL;

        qcow2_read_extensions(bs, header.header_length, ext_end,

                              &feature_table);

        report_unsupported_feature(bs, feature_table,

                                   s->incompatible_features &

                                   ~QCOW2_INCOMPAT_MASK);

        ret = -ENOTSUP;

        goto fail;

    }

    /* Check support for various header values */

    if (header.refcount_order != 4) {

        report_unsupported(bs, "%d bit reference counts",

                           1 << header.refcount_order);

        ret = -ENOTSUP;

        goto fail;

    }

    if (header.cluster_bits < MIN_CLUSTER_BITS ||

        header.cluster_bits > MAX_CLUSTER_BITS) {

        ret = -EINVAL;

        goto fail;

    }

    if (header.crypt_method > QCOW_CRYPT_AES) {

        ret = -EINVAL;

        goto fail;

    }

    s->crypt_method_header = header.crypt_method;

    if (s->crypt_method_header) {

        bs->encrypted = 1;

    }

    s->cluster_bits = header.cluster_bits;

    s->cluster_size = 1 << s->cluster_bits;

    s->cluster_sectors = 1 << (s->cluster_bits - 9);

    s->l2_bits = s->cluster_bits - 3; /* L2 is always one cluster */

    s->l2_size = 1 << s->l2_bits;

    bs->total_sectors = header.size / 512;

    s->csize_shift = (62 - (s->cluster_bits - 8));

    s->csize_mask = (1 << (s->cluster_bits - 8)) - 1;

    s->cluster_offset_mask = (1LL << s->csize_shift) - 1;

    s->refcount_table_offset = header.refcount_table_offset;

    s->refcount_table_size =

        header.refcount_table_clusters << (s->cluster_bits - 3);

    s->snapshots_offset = header.snapshots_offset;

    s->nb_snapshots = header.nb_snapshots;

    /* read the level 1 table */

    s->l1_size = header.l1_size;

    s->l1_vm_state_index = size_to_l1(s, header.size);

    /* the L1 table must contain at least enough entries to put

       header.size bytes */

    if (s->l1_size < s->l1_vm_state_index) {

        ret = -EINVAL;

        goto fail;

    }

    s->l1_table_offset = header.l1_table_offset;

    if (s->l1_size > 0) {

        s->l1_table = g_malloc0(

            align_offset(s->l1_size * sizeof(uint64_t), 512));

        ret = bdrv_pread(bs->file, s->l1_table_offset, s->l1_table,

                         s->l1_size * sizeof(uint64_t));

        if (ret < 0) {

            goto fail;

        }

        for(i = 0;i < s->l1_size; i++) {

            be64_to_cpus(&s->l1_table[i]);

        }

    }

    /* alloc L2 table/refcount block cache */

    s->l2_table_cache = qcow2_cache_create(bs, L2_CACHE_SIZE);

    s->refcount_block_cache = qcow2_cache_create(bs, REFCOUNT_CACHE_SIZE);

    s->cluster_cache = g_malloc(s->cluster_size);

    /* one more sector for decompressed data alignment */

    s->cluster_data = qemu_blockalign(bs, QCOW_MAX_CRYPT_CLUSTERS * s->cluster_size

                                  + 512);

    s->cluster_cache_offset = -1;

    s->flags = flags;

    ret = qcow2_refcount_init(bs);

    if (ret != 0) {

        goto fail;

    }

    QLIST_INIT(&s->cluster_allocs);

    /* read qcow2 extensions */

    if (qcow2_read_extensions(bs, header.header_length, ext_end, NULL)) {

        ret = -EINVAL;

        goto fail;

    }

    /* read the backing file name */

    if (header.backing_file_offset != 0) {

        len = header.backing_file_size;

        if (len > 1023) {

            len = 1023;

        }

        ret = bdrv_pread(bs->file, header.backing_file_offset,

                         bs->backing_file, len);

        if (ret < 0) {

            goto fail;

        }

        bs->backing_file[len] = '\0';

    }

    ret = qcow2_read_snapshots(bs);

    if (ret < 0) {

        goto fail;

    }

    /* Clear unknown autoclear feature bits */

    if (!bs->read_only && s->autoclear_features != 0) {

        s->autoclear_features = 0;

        ret = qcow2_update_header(bs);

        if (ret < 0) {

            goto fail;

        }

    }

    /* Initialise locks */

    qemu_co_mutex_init(&s->lock);

    /* Repair image if dirty */

    if (!(flags & BDRV_O_CHECK) && !bs->read_only &&

        (s->incompatible_features & QCOW2_INCOMPAT_DIRTY)) {

        BdrvCheckResult result = {0};

        ret = qcow2_check(bs, &result, BDRV_FIX_ERRORS);

        if (ret < 0) {

            goto fail;

        }

    }

    /* Enable lazy_refcounts according to image and command line options */

    opts = qemu_opts_create_nofail(&qcow2_runtime_opts);

    qemu_opts_absorb_qdict(opts, options, &local_err);

    if (error_is_set(&local_err)) {

        qerror_report_err(local_err);

        error_free(local_err);

        ret = -EINVAL;

        goto fail;

    }

    s->use_lazy_refcounts = qemu_opt_get_bool(opts, QCOW2_OPT_LAZY_REFCOUNTS,

        (s->compatible_features & QCOW2_COMPAT_LAZY_REFCOUNTS));

    qemu_opts_del(opts);

    if (s->use_lazy_refcounts && s->qcow_version < 3) {

        qerror_report(ERROR_CLASS_GENERIC_ERROR, "Lazy refcounts require "

            "a qcow2 image with at least qemu 1.1 compatibility level");

        ret = -EINVAL;

        goto fail;

    }

#ifdef DEBUG_ALLOC

    {

        BdrvCheckResult result = {0};

        qcow2_check_refcounts(bs, &result, 0);

    }

#endif

    return ret;

 fail:

    g_free(s->unknown_header_fields);

    cleanup_unknown_header_ext(bs);

    qcow2_free_snapshots(bs);

    qcow2_refcount_close(bs);

    g_free(s->l1_table);

    if (s->l2_table_cache) {

        qcow2_cache_destroy(bs, s->l2_table_cache);

    }

    g_free(s->cluster_cache);

    qemu_vfree(s->cluster_data);

    return ret;

}

static int qcow2_set_key(BlockDriverState *bs, const char *key)

{

    BDRVQcowState *s = bs->opaque;

    uint8_t keybuf[16];

    int len, i;

    memset(keybuf, 0, 16);

    len = strlen(key);

    if (len > 16)

        len = 16;

    /* XXX: we could compress the chars to 7 bits to increase

       entropy */

    for(i = 0;i < len;i++) {

        keybuf[i] = key[i];

    }

    s->crypt_method = s->crypt_method_header;

    if (AES_set_encrypt_key(keybuf, 128, &s->aes_encrypt_key) != 0)

        return -1;

    if (AES_set_decrypt_key(keybuf, 128, &s->aes_decrypt_key) != 0)

        return -1;

#if 0

    /* test */

    {

        uint8_t in[16];

        uint8_t out[16];

        uint8_t tmp[16];

        for(i=0;i<16;i++)

            in[i] = i;

        AES_encrypt(in, tmp, &s->aes_encrypt_key);

        AES_decrypt(tmp, out, &s->aes_decrypt_key);

        for(i = 0; i < 16; i++)

            printf(" %02x", tmp[i]);

        printf("\n");

        for(i = 0; i < 16; i++)

            printf(" %02x", out[i]);

        printf("\n");

    }

#endif

    return 0;

}

/* We have nothing to do for QCOW2 reopen, stubs just return

 * success */

static int qcow2_reopen_prepare(BDRVReopenState *state,

                                BlockReopenQueue *queue, Error **errp)

{

    return 0;

}

static int coroutine_fn qcow2_co_is_allocated(BlockDriverState *bs,

        int64_t sector_num, int nb_sectors, int *pnum)

{

    BDRVQcowState *s = bs->opaque;

    uint64_t cluster_offset;

    int ret;

    *pnum = nb_sectors;

    /* FIXME We can get errors here, but the bdrv_co_is_allocated interface

     * can't pass them on today */

    qemu_co_mutex_lock(&s->lock);

    ret = qcow2_get_cluster_offset(bs, sector_num << 9, pnum, &cluster_offset);

    qemu_co_mutex_unlock(&s->lock);

    if (ret < 0) {

        *pnum = 0;

    }

    return (cluster_offset != 0) || (ret == QCOW2_CLUSTER_ZERO);

}

/* handle reading after the end of the backing file */

int qcow2_backing_read1(BlockDriverState *bs, QEMUIOVector *qiov,

                  int64_t sector_num, int nb_sectors)

{

    int n1;

    if ((sector_num + nb_sectors) <= bs->total_sectors)

        return nb_sectors;

    if (sector_num >= bs->total_sectors)

        n1 = 0;

    else

        n1 = bs->total_sectors - sector_num;

    qemu_iovec_memset(qiov, 512 * n1, 0, 512 * (nb_sectors - n1));

    return n1;

}

static coroutine_fn int qcow2_co_readv(BlockDriverState *bs, int64_t sector_num,

                          int remaining_sectors, QEMUIOVector *qiov)

{

    BDRVQcowState *s = bs->opaque;

    int index_in_cluster, n1;

    int ret;

    int cur_nr_sectors; /* number of sectors in current iteration */

    uint64_t cluster_offset = 0;

    uint64_t bytes_done = 0;

    QEMUIOVector hd_qiov;

    uint8_t *cluster_data = NULL;

    qemu_iovec_init(&hd_qiov, qiov->niov);

    qemu_co_mutex_lock(&s->lock);

    while (remaining_sectors != 0) {

        /* prepare next request */

        cur_nr_sectors = remaining_sectors;

        if (s->crypt_method) {

            cur_nr_sectors = MIN(cur_nr_sectors,

                QCOW_MAX_CRYPT_CLUSTERS * s->cluster_sectors);

        }

        ret = qcow2_get_cluster_offset(bs, sector_num << 9,

            &cur_nr_sectors, &cluster_offset);

        if (ret < 0) {

            goto fail;

        }

        index_in_cluster = sector_num & (s->cluster_sectors - 1);

        qemu_iovec_reset(&hd_qiov);

        qemu_iovec_concat(&hd_qiov, qiov, bytes_done,

            cur_nr_sectors * 512);

        switch (ret) {

        case QCOW2_CLUSTER_UNALLOCATED:

            if (bs->backing_hd) {

                /* read from the base image */

                n1 = qcow2_backing_read1(bs->backing_hd, &hd_qiov,

                    sector_num, cur_nr_sectors);

                if (n1 > 0) {

                    BLKDBG_EVENT(bs->file, BLKDBG_READ_BACKING_AIO);

                    qemu_co_mutex_unlock(&s->lock);

                    ret = bdrv_co_readv(bs->backing_hd, sector_num,

                                        n1, &hd_qiov);

                    qemu_co_mutex_lock(&s->lock);

                    if (ret < 0) {

                        goto fail;

                    }

                }

            } else {

                /* Note: in this case, no need to wait */

                qemu_iovec_memset(&hd_qiov, 0, 0, 512 * cur_nr_sectors);

            }

            break;

        case QCOW2_CLUSTER_ZERO:

            qemu_iovec_memset(&hd_qiov, 0, 0, 512 * cur_nr_sectors);

            break;

        case QCOW2_CLUSTER_COMPRESSED:

            /* add AIO support for compressed blocks ? */

            ret = qcow2_decompress_cluster(bs, cluster_offset);

            if (ret < 0) {

                goto fail;

            }

            qemu_iovec_from_buf(&hd_qiov, 0,

                s->cluster_cache + index_in_cluster * 512,

                512 * cur_nr_sectors);

            break;

        case QCOW2_CLUSTER_NORMAL:

            if ((cluster_offset & 511) != 0) {

                ret = -EIO;

                goto fail;

            }

            if (s->crypt_method) {

                /*

                 * For encrypted images, read everything into a temporary

                 * contiguous buffer on which the AES functions can work.

                 */

                if (!cluster_data) {

                    cluster_data =

                        qemu_blockalign(bs, QCOW_MAX_CRYPT_CLUSTERS * s->cluster_size);

                }

                assert(cur_nr_sectors <=

                    QCOW_MAX_CRYPT_CLUSTERS * s->cluster_sectors);

                qemu_iovec_reset(&hd_qiov);

                qemu_iovec_add(&hd_qiov, cluster_data,

                    512 * cur_nr_sectors);

            }

            BLKDBG_EVENT(bs->file, BLKDBG_READ_AIO);

            qemu_co_mutex_unlock(&s->lock);

            ret = bdrv_co_readv(bs->file,

                                (cluster_offset >> 9) + index_in_cluster,

                                cur_nr_sectors, &hd_qiov);

            qemu_co_mutex_lock(&s->lock);

            if (ret < 0) {

                goto fail;

            }

            if (s->crypt_method) {

                qcow2_encrypt_sectors(s, sector_num,  cluster_data,

                    cluster_data, cur_nr_sectors, 0, &s->aes_decrypt_key);

                qemu_iovec_from_buf(qiov, bytes_done,

                    cluster_data, 512 * cur_nr_sectors);

            }

            break;

        default:

            g_assert_not_reached();

            ret = -EIO;

            goto fail;

        }

        remaining_sectors -= cur_nr_sectors;

        sector_num += cur_nr_sectors;

        bytes_done += cur_nr_sectors * 512;

    }

    ret = 0;

fail:

    qemu_co_mutex_unlock(&s->lock);

    qemu_iovec_destroy(&hd_qiov);

    qemu_vfree(cluster_data);

    return ret;

}

static coroutine_fn int qcow2_co_writev(BlockDriverState *bs,

                           int64_t sector_num,

                           int remaining_sectors,

                           QEMUIOVector *qiov)

{

    BDRVQcowState *s = bs->opaque;

    int index_in_cluster;

    int n_end;

    int ret;

    int cur_nr_sectors; /* number of sectors in current iteration */

    uint64_t cluster_offset;

    QEMUIOVector hd_qiov;

    uint64_t bytes_done = 0;

    uint8_t *cluster_data = NULL;

    QCowL2Meta *l2meta = NULL;

    trace_qcow2_writev_start_req(qemu_coroutine_self(), sector_num,

                                 remaining_sectors);

    qemu_iovec_init(&hd_qiov, qiov->niov);

    s->cluster_cache_offset = -1; /* disable compressed cache */

    qemu_co_mutex_lock(&s->lock);

    while (remaining_sectors != 0) {

        l2meta = NULL;

        trace_qcow2_writev_start_part(qemu_coroutine_self());

        index_in_cluster = sector_num & (s->cluster_sectors - 1);

        n_end = index_in_cluster + remaining_sectors;

        if (s->crypt_method &&

            n_end > QCOW_MAX_CRYPT_CLUSTERS * s->cluster_sectors) {

            n_end = QCOW_MAX_CRYPT_CLUSTERS * s->cluster_sectors;

        }

        ret = qcow2_alloc_cluster_offset(bs, sector_num << 9,

            index_in_cluster, n_end, &cur_nr_sectors, &cluster_offset, &l2meta);

        if (ret < 0) {

            goto fail;

        }

        assert((cluster_offset & 511) == 0);

        qemu_iovec_reset(&hd_qiov);

        qemu_iovec_concat(&hd_qiov, qiov, bytes_done,

            cur_nr_sectors * 512);

        if (s->crypt_method) {

            if (!cluster_data) {

                cluster_data = qemu_blockalign(bs, QCOW_MAX_CRYPT_CLUSTERS *

                                                 s->cluster_size);

            }

            assert(hd_qiov.size <=

                   QCOW_MAX_CRYPT_CLUSTERS * s->cluster_size);

            qemu_iovec_to_buf(&hd_qiov, 0, cluster_data, hd_qiov.size);

            qcow2_encrypt_sectors(s, sector_num, cluster_data,

                cluster_data, cur_nr_sectors, 1, &s->aes_encrypt_key);

            qemu_iovec_reset(&hd_qiov);

            qemu_iovec_add(&hd_qiov, cluster_data,

                cur_nr_sectors * 512);

        }

        qemu_co_mutex_unlock(&s->lock);

        BLKDBG_EVENT(bs->file, BLKDBG_WRITE_AIO);

        trace_qcow2_writev_data(qemu_coroutine_self(),

                                (cluster_offset >> 9) + index_in_cluster);

        ret = bdrv_co_writev(bs->file,

                             (cluster_offset >> 9) + index_in_cluster,

                             cur_nr_sectors, &hd_qiov);

        qemu_co_mutex_lock(&s->lock);

        if (ret < 0) {

            goto fail;

        }

        while (l2meta != NULL) {

            QCowL2Meta *next;

            ret = qcow2_alloc_cluster_link_l2(bs, l2meta);

            if (ret < 0) {

                goto fail;

            }

            /* Take the request off the list of running requests */

            if (l2meta->nb_clusters != 0) {

                QLIST_REMOVE(l2meta, next_in_flight);

            }

            qemu_co_queue_restart_all(&l2meta->dependent_requests);

            next = l2meta->next;

            g_free(l2meta);

            l2meta = next;

        }

        remaining_sectors -= cur_nr_sectors;

        sector_num += cur_nr_sectors;

        bytes_done += cur_nr_sectors * 512;

        trace_qcow2_writev_done_part(qemu_coroutine_self(), cur_nr_sectors);

    }

    ret = 0;

fail:

    qemu_co_mutex_unlock(&s->lock);

    while (l2meta != NULL) {

        QCowL2Meta *next;

        if (l2meta->nb_clusters != 0) {

            QLIST_REMOVE(l2meta, next_in_flight);

        }

        qemu_co_queue_restart_all(&l2meta->dependent_requests);

        next = l2meta->next;

        g_free(l2meta);

        l2meta = next;

    }

    qemu_iovec_destroy(&hd_qiov);

    qemu_vfree(cluster_data);

    trace_qcow2_writev_done_req(qemu_coroutine_self(), ret);

    return ret;

}